Skip to content
Back to work

Professor-intelligence MCP server · live APIs

ProfGraph

An LLM can't answer 'is this section worth taking?' without live ratings and real grade distributions.

View source

What I built

A Python MCP server exposing 10 LLM-callable tools over the RateMyProfessors GraphQL API and UT Dallas's Nebula grade API, with a regex NLP teaching-style classifier, GPA-conditioned grade prediction, and a Starlette REST/OpenAPI bridge. Verified end to end against live APIs (real ratings + 22-semester grade distributions) and built to degrade gracefully when an upstream field changes.

Stack

PythonMCPGraphQLStarletteREST/OpenAPI

Status

Shipped; verified returning live data. 54 unit tests + 13/13 MCP e2e.